R is better than MATLAB

Import data

Warning: package 'dplyr' was built under R version 3.4.1

Will analyze data broken down by the following variables

condtnVariableNames = c('subject','wordEccentricity','side')

Fit mixture model with R (if fitting not previously done and cached).

# A tibble: 6 x 8
# Groups:   subject, wordEccentricity, side [6]
  subject wordEccentricity  side   efficacy   latency precision      val
   <fctr>            <dbl> <chr>      <dbl>     <dbl>     <dbl>    <dbl>
1      T1              0.8  left 0.80708946 0.4834758 0.8188051 201.1395
2      T1              0.8 right 0.67787890 0.5979391 0.8119572 237.4400
3      T1              6.0  left 0.41880287 0.9465679 0.4632598 278.3538
4      T1              6.0 right 0.72873293 0.6843011 0.5726754 204.7284
5     T10              0.8  left 0.08519846 0.1253751 0.6781913 347.4345
6     T10              0.8 right 0.20795796 0.7397509 0.6158247 335.4381
# ... with 1 more variables: warnings <fctr>

Plot some of the fits

• yellow = guessing component
• light blue = Gaussian component
• green = sum of the guessing and Gaussian components. In other words, the histogram heights predicted by the model
• dark blue = continuous Gaussian. This helps get a sense of the effect of discretising the Gaussian. For instance, it’s possible (especially using Pat’s method, it seems), for the Gaussian peak to fly high above the bars and still fit the discrete bins (or bin centers, in Pat’s method), suggesting an undesirably high estimates of the efficacy (likely accompanied by an undesirably low precision)

Warning: Removed 19 rows containing non-finite values (stat_bin).